Service Selection With QoS Correlations in Distributed Service-Based Systems

• Main Authors: Dongwei Li, Dayong Ye, Nan Gao, Shuliang Wang
• Format: Article
• Language: English
• Published: IEEE 2019-01-01
• Series: IEEE Access
• Subjects: Distributed service-based systems; service selection; QoS correlations; user requirements
• Online Access: https://ieeexplore.ieee.org/document/8752211/

Service selection is an important research problem in distributed service-based systems, which aims to select proper services to meet user requirements. A number of service selection approaches have been proposed in recent years. Most of them, however, overlook quality-of-service (QoS) correlations, which broadly exist in distributed service-based systems. The concept of QoS correlations involves two aspects: 1) QoS correlations among services and 2) QoS correlations of user requirements. The first aspect means that some QoS attributes of service not only depend on the service itself but also have correlations with other services, e.g., buying service 1 and then getting service 2 with half price. The second aspect means the relationships among QoS attributes of user requirements, e.g., a user can accept a service with fast response time and high service cost or the user can also accept a service with slow response time and low service cost (Fig. 1). These correlations significantly affect user selection of services. Currently, only a few existing approaches have considered QoS correlations among services, i.e., the first aspect, but they still overlook QoS correlations of user requirements, i.e., the second aspect, which are also very important in distributed service-based systems. In this paper, a novel service selection approach is proposed, which not only considers QoS correlations of services but also accounts for QoS correlations of user requirements. This approach, to the best of our knowledge, is the first one which considers QoS correlations of user requirements. Also, this approach is decentralized which can avoid the single point of failure. The experimental results demonstrate the effectiveness of the proposed approach.